© Neuronpedia 2026
    Privacy & TermsBlogGitHubSlackTwitterContact
    Neuronpedia logo - a computer chip with a rounded viewfinder border around it

    Neuronpedia

    Jacobian LensNEW
    Natural Language
    Autoencoders
    NEW    Assistant AxisNEWCircuit TracerUPDATESteerSAE EvalsExportsinterp-engineNEWAPI Community BlogPrivacy & TermsContact
    1. Home
    3. Gemma-2-27B
    5. 10-GEMMASCOPE-RES-131K
    7. 22442
    Prev
    Next
    INDEX
    Explanations

    begins of words

    np_acts-logits-general · gemini-2.5-flash-lite

    The neuron fires on subword tokens that occur at the very beginning of longer words (i.e. word-initial WordPiece prefixes).

    oai_token-act-pair · o4-miniTriggered by @jyhe0408
    New Auto-Interp
    Top Features by Cosine Similarity
    Configuration
    google/gemma-scope-27b-pt-res/layer_10/width_131k
    Prompts (Dashboard)
    24,576 prompts, 128 tokens each
    Dataset (Dashboard)
    monology/pile-uncopyrighted
    No Configuration Found
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Negative Logits
    诘
    -2.88
    贔
    -2.83
    汌
    -2.72
     которы
    -2.66
    摄像
    -2.48
    飚
    -2.47
    苹
    -2.41
    栒
    -2.39
    ING
    -2.38
    迳
    -2.36
    POSITIVE LOGITS
    ’
    3.95
    This
    3.48
    ed
    3.45
    は
    3.05
    These
    2.86
     This
    2.80
    8
    2.64
    1
    2.61
    er
    2.53
    4
    2.52
    Activations Density 0.025%

    No Known Activations